Mortality associated with oral contraceptive use: 25 year follow up of cohort of 46 000 women from Royal College of General Practitioners' oral contraception study

BMJ 1999; 318 doi: http://dx.doi.org/10.1136/bmj.318.7176.96 (Published 9 January 1999)
Cite this as: BMJ 1999;318:96
  1. Valerie Beral, professor (beral{at}icrf.icnet.uk)a,
  2. Carol Hermon, scientific officera,
  3. Clifford Kay, directorb,
  4. Philip Hannaford, professorb,
  5. Sarah Darby, professora,
  6. Gillian Reeves, statistician.a
  1. aImperial Cancer Research Fund, Cancer Epidemiology Unit, Radcliffe Infirmary, Oxford OX2 6HE
  2. bRoyal College of General Practitioners, Manchester Research Unit, Parkway House, Manchester M22 4DB
  1. Correspondence to: Professor Beral
  • Accepted 15 September 1998

Abstract

Objective: To describe the long term effects of the use of oral contraceptives on mortality.

Design: Cohort study with 25year follow up. Details of oral contraceptive use and of morbidity and mortality were reported six monthly by general practitioners. 75% of the original cohort was “flagged” on the NHS central registers.

Setting: 1400 general practices throughout Britain.

Subjects: 46000 women, half of whom were using oral contraceptives at recruitment in 1968-9.Median age at end of follow up was 49 years.

Main outcome measures: Relative risks of death adjusted for age, parity, social class, and smoking.

Results: Over the 25year follow up 1599deaths were reported. Over the entire period of follow up the risk of death from all causes was similar in ever users and never users of oral contraceptives (relative risk=1.0, 95% confidence interval 0.9 to 1.1; P=0.7) and the risk of death for most specific causes did not differ significantly in the two groups. However, among current and recent (within 10years) users the relative risk of death from ovarian cancer was 0.2(0.1to 0.8; P=0.01), from cervical cancer 2.5(1.1to 6.1; P=0.04), and from cerebrovascular disease 1.9(1.2to 3.1,P=0.009). By contrast, for women who had stopped use >=10 years previously there were no significant excesses or deficits either overall or for any specific cause of death.

Conclusion: Oral contraceptives seem to have their main effect on mortality while they are being used and in the 10years after use ceases. Ten or more years after use ceases mortality in past users is similar to that in never users.

Editorial by Skegg

Key messages

  • This 25year follow up of 46000 UK women found a decrease in mortality from ovarian cancer and an increase in mortality from circulatory diseases and cervical cancer among women were using oral contraceptives or had used them in the past 10years

  • 10 or more years after stopping use mortality was similar in pas users and never users

  • Oral contraceptives seem to have their main effect on mortality mainly while they are being used and in the 10years after stopping use

  • There is little evidence to suggest any persistent adverse effect 10or more years after use of oral contraceptives ceases

Introduction

Oral contraceptives have been available for 40years and, although their short term effects on health have been studied in detail, 1 2 comparatively little is known about whether these effects persist after use stops. The Royal College of General Practitioners' oral contraception study was set up in 1968to monitor the health of women who had used oral contraceptives. We present the results of a 25year follow up of that population examining the effect of use of oral contraceptives on mortality in the long term.

Subjects and methods

Over 14months from May 1968,1400general practitioners throughout the United Kingdom recruited 23000 women who were using oral contraceptives and a similar number who had never used them into the oral contraception study.1 Most women (98%) were white and all were married or living as married. General practitioners were asked to provide information on oral contraceptives prescribed, pregnancies, new illness, or death for each subject every six months. During the early years of the study some general practitioners withdrew their patients and some women moved and left the study. In 1976-7an attempt was made to “flag” the cohort on the NHS central registers in Southport and Edinburgh to provide information on death and cause of death for women who were no longer being followed regularly by their general practitioner. About 75% of the original cohort was successfully flagged, and these women have been followed for death and emigration since then. The remaining 25% could not be flagged because they or their general practitioners had left the study before the flagging procedure could be instigated and the personal details required for flagging were not available to the investigators. The mortality of the women who were followed regularly by their general practitioner was similar to that of women who left the study.3

This analysis includes deaths up to 31December 1993.We obtained a copy of the death certificate for all women who had died, and CK or PH coded the cause of death according to ICD-8 (international classification of diseases, eighth revision),4 occasionally supplementing information from the death certificate with details provided by general practitioners.5 Person-years of follow up were calculated from the date of recruitment up to the date of death for the 1599women who had died, up to the date of last contact with the general practitioner for 10958 women who were not flagged on the NHS registers, or up to 31December 1993for 33554 women who were flagged on the NHS registers and alive on that date.3 For women who were no longer being followed by their general practitioner before 1January 1977but were flagged no person-years were included for the period between the date of last contact with their general practitioner and 1January 1977because the ascertainment of deaths during that period may have been incomplete.3

Person-years were categorised by age (16-19,20-24…70-79), parity (0,1-2,>=3, not known), social class at recruitment (I-II, III, IV-V, other), and cigarette smoking at recruitment (0,1-14/day, >=15/day, not known) with a standard computer program.6 Person-years were further subdivided according to whether the women had taken oral contraceptives and, where appropriate, by duration of use and time since first and last use. At recruitment half (23000) of the subjects were using oral contraceptives, but by 31December 1993,63% had used them at some time. Women who started using oral contraceptives after recruitment contributed person-years to the “never user” category up to the date that they began using them. For women who were flagged on the NHS registers but no longer regularly followed up by their general practitioner, we assumed that past users and never users who were over the age of 40at the date of last contact did not subsequently take oral contraceptives. For current users and women aged under 40at the time of last contact with their general practitioners, we assumed that use continued for two years as stated at the time of last contact, but thereafter use was classified as unknown. These assumptions about subsequent use of oral contraceptives are similar to those used in analyses of other cohort studies.7

The results presented here are based on 853517 person-years of follow up until 31December 1993: 517519 in women who had used oral contraceptives and 335998 in women who had never used them. Standardised mortality ratios were calculated by using mortality for women in England and Wales as a standard.6 Relative risks were adjusted for age, parity, social class, and smoking with the Poisson regression program module of EPICURE.8 P values are two tailed.

Results

By 31December 1993the cohort had been followed for 25years and the median age of the women was 49years (48for ever users of oral contraception and 50for never users). During that period 1599deaths were reported, 945in ever users and 654in never users (table 1). The death rate from all causes combined was 21% lower than in the UK population (overall standardised mortality ratio=79). The relative risk of death from all causes combined after adjustment for age, parity, social class, and cigarette smoking did not differ significantly between ever users and never users (relative risk=1.0, 95% confidence interval 0.9to 1.1; P=0.7).

View this table:
Table 1

Standardised mortality ratios in ever users and never users of oral contraceptives and relative risk in ever users compared with never users

Table 1 also shows standardised mortality ratios and adjusted relative risks of death for common specific causes and groups of causes of death (and also for some particular causes that have been reported to be affected by oral contraceptive use) according to ever use of oral contraceptives. For most specific causes of death the standardised mortality ratios in ever users and never users of oral contraceptives were around 100and did not differ significantly between the two groups. The few exceptions were colorectal cancer and ovarian cancer, for which the relative risks of death in users were significantly below 1.0,and cerebrovascular disease and all violent and accidental causes of death, for which the relative risks were significantly greater than 1.0.Ever use is, however, a crude measure of use of oral contraceptives.

Table 2 shows for various causes the relative risk of death compared with never users in relation to the number of years since oral contraceptives were first used. Within the first 10 years of starting use of oral contraceptives there was a significant excess mortality from all causes of death (relative risk=1.2, 95% confidence interval 1.0to 1.50; P=0.03), all circulatory diseases (2.2,1.5to 3.2; P<0.0001), and cerebrovascular disease (2.7,1.5to 4.9; P=0.0008). However, the excess mortality from these causes fell with time, this trend being significant for all circulatory disease (P=0.002) and cerebrovascular disease (P=0.02). There were 380deaths in women who began using oral contraceptives more than 20years before the end of follow up, and this group showed no significant excess or deficit in mortality from any specific condition or overall.

View this table:
Table 2

Relative risk of death in users of oral contraceptives compared with never users according to time since first use

Table 3 shows the pattern of risk of death for various conditions in relation to the time since stopping use of oral contraceptives. By the end of follow up the median time since last use in the cohort was 17years. Significant increases or decreases in risk were found mainly in current users or those who had stopped use within the past 10years—for example, women who were current users or who stopped use in the past five years had a significantly reduced risk of ovarian cancer (0.1,0.0to 0.9; P=0.04) and a significant excess of all circulatory diseases (1.7,1.2to 2.4; P=0.006) and cerebrovascular disease (1.9,1.1to 3.4; P=0.03) and women who had stopped use five to nine years previously had an significant excess risk of cervical cancer (3.0,1.1to 8.1; P=0.03) and cerebrovascular disease (2.0,1.1to 3.7; P=0.02). Among women who had stopped use 15or more years previously most of the relative risks were around 1.0.For ovarian cancer there was a weak suggestion that the protective effect associated with current or recent use wore off (test for trend, P=0.05).

View this table:
Table 3

Relative risk of death in users of oral contraceptives compared with never users according to time since last use

Among ever users of oral contraceptives, the average duration of use was five years. Table 4 shows the relative risk of death in relation to the duration of use of oral contraceptives. Women who used oral contraceptives for 10or more years had a significant excess mortality from lung cancer (2.0,1.1to 3.5; P=0.02) and cervical cancer (4.1,1.6to 10.6; P=0.003). The excess deaths from lung cancer were mainly among smokers (17deaths in smokers and three in non-smokers), the relative risk associated with 10 or more years of use of oral contraceptives being 2.0for smokers and 2.2for non-smokers. This excess may be a chance finding or perhaps due to residual confounding. There was also a significant trend of increasing mortality for all cancers combined and for cervical cancer in relation to duration of use (P=0.02 and 0.03, respectively).

View this table:
Table 4

Relative risk of death in users of oral contraceptives compared with never users according to duration of use

Duration of use and time since first and last use of oral contraceptives were highly correlated, with current and recent users being more likely to have used contraceptives for longer. Table 5 shows the relative risk of death among ever users of oral contraceptives according to time since last use of oral contraceptives and duration of use. All significant results were confined to women currently using oral contraceptives or who had stopped in the past 10years, although among such women duration of use was not associated with a significant increase or decrease in mortality from any particular cause or overall. Women who stopped using oral contraceptives 10or more years previously had no significant increases or decreases in relative risk of death from any cause, even if they had used them for 10years or more. There were, however, only 54deaths in this subgroup.

View this table:
Table 5

Relative risk† of death in users of oral contraceptives compared with never users according to time since last use and duration of use

Discussion

Our results suggest that most of the effects of oral contraceptives on mortality occur in current or recent users and that few, if any, effects persist 10years after stopping use. These results relate predominately to use of combined oral contraceptives containing 50µg oestrogen.1

Information on use of oral contraceptives was recorded prospectively at six monthly intervals by the subjects' general practitioner and so is unlikely to be biased by subsequent events. Furthermore, because three quarters of the original cohort was “flagged” on the NHS central registers in England and Scotland and so followed routinely for death, the findings are likely to be representative of the majority of the women originally recruited. Mortality was similar in women who remained under regular follow up by their general practitioner and in women who did not.3 That overall mortality in our cohort was about 20% below the national average is not unexpected since women with severe chronic illnesses were not recruited. 1 3

Death certificates were obtained for all women who died. There was good agreement between cause of death recorded on the death certificate and that reported by general practitioners.5 We adjusted for the potential confounding factors of age, parity, social class, and cigarette smoking. Information on age and parity was updated throughout the follow up, whereas social class and smoking details were recorded at entry only. Information on subsequent smoking habits was obtained in 1994-5for 11797 members of the original cohort; re-estimation of the risk of myocardial infarction associated with oral contraceptive use based on the updated data gave virtually identical results to those based on smoking history at entry.9 Use of information on smoking at entry is thus unlikely to have biased our results. We did not adjust for hypertension or other heart disease because such conditions could be in the causal pathway for death from circulatory diseases. No data on family history of these conditions or of cancer were available, but the absence of such information is unlikely to produce spurious associations suggesting that mortality varies according to the timing of oral contraceptive use.

The specific diseases showing significant excesses or deficits in mortality in our study were generally consistent with the results of other studies on the incidence of these diseases. 1 2 10 Other cohort studies have reported no significant changes in mortality among women who have ever used oral contraceptives, which might at first sight be interpreted as inconsistent with their known effects on incidence of disease. 11 12 What our results highlight, however, is that the effects of oral contraceptives on mortality occur mainly in current and recent users.

The effects of oral contraceptives on circulatory diseases are already recognised to be largely confined to current users, especially if they also smoke.1316 There has been concern, however, that oral contraceptive use might affect risk of cancer many years after use stops. The collaborative reanalysis of the worldwide data on the relation between breast cancer and oral contraceptive use, which included data from this study, showed that the incidence of breast cancer was slightly increased while women used oral contraceptives and in the 10years after stopping use but that there was no excess risk 10or more years after stopping.7 Our results are consistent with this finding and suggest that other cancers of the female reproductive organs may also be affected by current and recent use of oral contraceptives but may wear off after use stops. The number of deaths from each type of cancer was small, and further data are needed to confirm our findings. Continued follow up of this and other cohorts will yield important information for the many millions of women throughout the world who have used oral contraceptives.

Acknowledgments

We thank the 1400doctors who have contributed data to the study.

Contributors: CK set up the oral contraception study and PH took over as director in 1994.CH, SD, GR, and VB contributed to the data analysis. VB prepared the first draft of the manuscript and all other authors have contributed to it. CK is guarantor for the quality of the data; VB and CH are guarantors for the analyses and text.

Funding: Royal College of General Practitioners, British Heart Foundation, Imperial Cancer Research Fund, Medical Insurance Agency, Medical Research Council, Schering AG (Berlin), Schering Health Care (UK), Searle, and Wyeth-Ayerst International (USA).

Competing interests: None declared.

References

THIS WEEK'S POLL